See axolotl config

axolotl version: 0.11.0.dev0

base_model: AlexHung29629/Magistral-Small-2506
tokenizer_use_mistral_common: false

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_fused_linear_cross_entropy: true

unfrozen_parameters:
  - ^lm_head.+$
  - ^.+embed_tokens.+$

datasets:
  - path: AlexHung29629/rr-mg-segmented
    type: input_output
remove_unused_columns: false

sequence_len: 8192
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true

use_tensorboard: true

save_only_model: true
gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 4
optimizer: adamw_torch_fused
#optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 1e-5
max_grad_norm: 1.0

adam_beta1: 0.9
adam_beta2: 0.95
adam_epsilon: 1e-8

bf16: true
tf32: false

warmup_ratio: 0.05
saves_per_epoch: 1
weight_decay: 0


train_on_inputs: false

flash_attention: true
#deepspeed: /workspace/output/zero3.json

fsdp:
  - full_shard
  - auto_wrap

fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: false
  fsdp_use_orig_params: true
  fsdp_cpu_ram_efficient_loading: true
  fsdp_activation_checkpointing: true
  fsdp_transformer_layer_cls_to_wrap: MistralDecoderLayer
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
seed: 42

hub_model_id: AlexHung29629/fix_magistra4
output_dir: /workspace/output/output_dir4
dataset_processes: 0
dataset_prepared_path: /workspace/output/dataset_prepared
torch_compile: false

added_tokens_overrides:
  32: "[ARGS]"
  33: "[CALL_ID]"
  

fix_magistra4

This model is a fine-tuned version of AlexHung29629/Magistral-Small-2506 on the AlexHung29629/rr-mg-segmented dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 8
• total_eval_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 7
• training_steps: 140

Training results

Framework versions

• Transformers 4.52.4
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1